Re: Air Force helping to undermine Ada

[srctran@world.std.com (Gregory Aharonian)]

>DoD does not care if Xerox saves money or not.  Therefore there is no
>reason for them to buy a bunch of object code that was created from
>Ada rather than a bunch created from C, or C++, or BRUIN.
>
>I personally think we are heading for a slow crisis because of the
>mass of incomprehendable C that so many vendors are staking their
>business upon.  There will come a time when the pile is no longer
>maintainable, and the company will be in trouble.  Lotus went through
>a crisis like this a few years ago, and the telephone companies are
>going through it right now.  Proper use of Ada would help avoid this.

Bob,
	This is wishful thinking on your part.  When it comes to spending
their OWN money in the non-mandated world, people are not spending it on
Ada.  You have been on the dole too long to know what it is to make a
decision about spending your own money.  I agree Ada is probably a better
language technically.  But that is not the sole criteria for choosing a
language; there are a variety of socioeconomic factors that the DoD
continues to ignore.  But these arguments about Ada and C/C++ are like
the old arguments about democracy versus communism.  The only thing that
ever really counted was and is which direction people were choosing with
their own feet.  Similarly, all that really counts is how people spend
their own money, and it is rarely on Ada.  Eventually the Ada world will
be so far behind the non-mandated world, the DoD will have no choice but
to drop the mandate in order to fulfull its mission, as the supply base of
Ada programmers and tools won't be large enough (as the FAA reports allude
to to some extent).

>Greg, I have to wonder why you keep attacking STARS.  You never seem
>to know much about the program, but you sure have an opinion of it! 
>I can only assume that your opinion is based on the same kind of logic
>as that you demonstrated with the Xerox example.

Bob,
      Which version of STARS are you referring to?  The program has changed
so many times its hard to keep a steady aim.  But in general, here are some
reasons I attack STARS.
 
1)  ONE OF THE STARS GOALS WAS A LIE
 
    During the initial stages of the STARS program (the pre-RFP, the RFP,
the initial efforts, and promises to the press and Congress) one of the
often stated goals was that STARS would increase software productivity for
the DoD by a factor of 10 (TEN).  During the RFP stage, and for a time
after, a few of us complained that never before had any project anywhere
increased productivity by a factor of ten in software development, and that
the STARS program had no sub-project to build a economic model to assess
such a claim and collect the relevant data.  Yet the STARS program stuck
to this outrageous claim.  About a year into the program, spring of 1990, I
had a phone conversation with the then program manager, Colonel Joseph Greene,
during which I asked how the STARS program was going to validate such a claim,
especially since hardware and compiler advances alone were boosting
productivity.  He said, and I quote, "We are assuming that the gains in
productivity will be so obvious we won't have to measure them".
    Finally, in 1991, the STARS program tasked a group at the IDA to look
at the problem.  Their report, given at the 1991 STARS conference concluded
that: "Though the model and its parameters are subject to change, preliminary
conclusions are: small increases in the speed of technology adoption result
in large savings, the STARS program appears to be cost-effective, and
achieving a decrease by a factor of two in software expenditures by the
year 2000 WILL BE DIFFICULT.
 
   Thus, one of the things I know is that one of the main initial goals of
STARS was a lie.
 
 
2)  STARS CONTINUES TO IGNORE VHDL
 
    Defense weapon systems consist of two, ever overlapping items: hardware
and software.  To standardize software design and maintenance, the DoD
developed Ada.  To standardized hardware design and maintenance, the DoD
developed VHDL. Both are fine standards with a very close syntactic structure.
Thus you would figure that the DoD look into every possible way to integrate
Ada and VHDL, which would allow the closer integration of hardware and
software activities.  Certainly a better meta idea.
    Yet the STARS program, the ideal environment for such an exploration,
has completely ignored use of VHDL, even as VHDL is sweeping the commercial
digital design world, and now the Air Force is seeking proposals for an
analog VHDL - AHDL.  VHDL has been so ignored by the DoD that most VHDL
tool suppliers only have their products generate C code from a VHDL
specification, when it would be easier to generate Ada code, given the
syntactic similarities between the languages  ( I doubt highly anyone with
STARS has ever compared the Ada and VHDL grammars to discover this ).
 
   Thus, another thing I know is that STARS lack of vision has prevented
it from helping to build bridges from Ada to VHDL.
 
 
3)  STARS IS EMBARASSED ABOUT ITS RESULTS
 
    I am always hearing about the great technology and breakthroughs coming
out of the STARS program, with the emphasis on 'hearing'.  I go to many of
the commercial software engineering conferences and trade shows each year,
such as Object World, CASE World, Software Engineering, Embedded Systems,
and NEVER do I see any of the STARS technology being talked about on or
display.  These perfect opportunities to present STARS technology to the
non-Mandated world in order to get people interested in STARS and Ada are
ignored.  These perfect opportunities to present STARS technology at a
booth in order to be discovered by small, new, software engineering companies
who could contribute to STARS are ignored.  These perfect opportunities to
present STARS technology to the general public to get feedback and critiques
from people not on the DoD dole are ignored.
    In short, the STARS program has consciously ignored presenting itself
to the general public, at a great disservice both to the general programming
community and the DoD.  One can only assume that the STARS program is
embarassed by its results, especially groups like IBM, which offers totally
different technology to its commercial customers for solving the same problem.
    After all, when an Air Force unit can develop a CASE tool and get a US
patent for it, and we still don't hear much pubicly about STARS, you have to
wonder how embarassing these self-labeled breakthroughs really are.
 
    Thus, another thing I know about STARS is that the quality of the results
is so low that everyone is afraid to present this technology in public forums.
 
 
4)  STARS IS EMBARRASSED BY ITSELF
 
    As I have mentioned on comp.lang.ada, there are a variety of new industry
groups in software engineering to promote new software technologies.  For
example, one such group is the Object Management Group, set up to coordinate
object oriented programming and software tools.  Few, if any, of the STARS
contractors are members of this group and show up at Object World.  Since it
does not cost much to join the group, compared to the profits made on STARS
contracts, it is hard to come up with a reason that STARS contractors are
not members, other than their having an isolationist tendency.
 
    Thus another thing I know is that STARS for the most part is rejecting
contacts with the outside world.
 
5)  THE STARS REPOSITORY EFFORT ASSET IS A FAILURE
 
    I would like to say that most people think that ASSET is a joke, but
that would require that most people know much about ASSET.  Here is a software
repository staffed by people with no background in software repositories, not
doing any of things other established software reposities are doing.  They
don't post FAQs to comp.lang.ada, they are not anonymous-ftp accessible,
they are not mail server accessible, they publish no marketing newsletters,
they do not use CDROMs, their schema for components is useless for commercial
suppliers of reusable software, they have no active program for seeking out
resuable Ada software, they have little if any contact with non-DoD federal
software repositories to seek advice, they reject the advice of commercial
maintainers of software repositories and reusable software.
    Even though I maintain the largest database of information on reusable
Defense software on a budget of a few thousand dollars a year, not once has
anyone from ASSET ever called to ask how I do what I do, in order to help
their own operations.  Simply they don't care about running something more
than a sit-on-your-butt passive repository.
 
(One other thing.  As part of STARS, IDA developed a spreadsheet economic
model to make the above assessment that reducing costs by the year 2000 will
be difficult.  Not once has IDA ever bothered or cared to post information to
comp.lang.ada, or anywhere else, about this spreadsheet and its availability.
Like many other developed software items as part of STARS, the incompetency of
promoting its reuse is ongoing and criminal.)
 
    In short, STARS' ASSET software repository is a failure, its people
unqualified to run a repository, and all of this being tolerated by the
current STARS program office (partly because no one in the STARS program
office has experience running a repository, especially with their own 
money.  The blind leading the blind).
 
 
==============================================================================
 
 
    Thus, before you get on my case, please get on your own.  If my critiques
are not well developed or thought out, it is only because I do not have your
resources and funding.  You try running STARS on a budget of a few thousand
dollars a year, and see how far you get.
 
Greg Aharonian
Source Translation & Optimization
-- 
**************************************************************************
Greg Aharonian
Source Translation & Optimiztion
P.O. Box 404, Belmont, MA 02178